In a team-based competitive game, agents cooperate to enhance their collective performance in winning the game. An interesting research problem in a team-based game is the role assignment problem (RAP). The problem requires agents to decide their respective roles based on real-time feedback from a dynamically changing environment. The Minority Game (MG), as used in modeling financial marketing problems, has shown similar characteristics that meet the fundamental requirements of RAP. In this paper, we propose a formulation of MG strategies for studying RAP in a specific team-based game: RoboCup Simulation League (RSL). Through experimentation, we demonstrate that MG strategies improve the effectiveness of role assignment among agents. The improvement validates some characteristics, e.g., the phase transition phenomenon on the memory size, as discovered in the theoretical MG model.